# RE: st: gologit2

 From Richard Williams To statalist@hsphsun2.harvard.edu, statalist@hsphsun2.harvard.edu Subject RE: st: gologit2 Date Thu, 17 Apr 2008 22:53:34 -0500

```At 08:14 PM 4/17/2008, David Jacobs wrote:
```
A student and I have about 1300 U.S. state-years in a pooled time series analysis of a state legal outcome that is measured as an ordinal scale (I plan to cluster on the state IDs to adjust for the pooled nature of the data or to use the pooled ordinal estimators in Limdep if I have to).

I understand, of course, how to use the BIC test to compare models, but I don't understand how this test can be used to test for the absence of proportionality in an ordinal logit of probit analysis.
Here is an example. You need gologit2, available from SSC:

. use "http://www.indiana.edu/~jslsoc/stata/spex_data/ordwarm2.dta";
(77 & 89 General Social Survey)

. quietly ologit warm yr89 male white age ed prst

. est store proportional

. quietly gologit2 warm yr89 male white age ed prst

. est store nonproportional

. lrtest proportional nonproportional, stats force

Likelihood-ratio test LR chi2(12) = 49.20
(Assumption: proportional nested in nonproportio~l) Prob > chi2 = 0.0000

-----------------------------------------------------------------------------
Model | Obs ll(null) ll(model) df AIC BIC
-------------+---------------------------------------------------------------
proportional | 2293 -2995.77 -2844.912 9 5707.825 5759.463
nonproport~l | 2293 -2995.77 -2820.311 21 5682.622 5803.112
-----------------------------------------------------------------------------
Note: N=Obs used in calculating BIC; see [R] BIC note

The likelihood ratio test says to reject proportional odds. The BIC test likes proportional odds better. I guess that makes the AIC test the tiebreaker, and it likes nonproportional odds better. If you use gologit2's -autofit- option, you can find an intermediate model that fits best of all. For more on gologit2, see

http://www.nd.edu/~rwilliam/gologit2/index.html

By the way, I can't get slogit to work at all (the Stata rountine won't give estimates) perhaps (?) because we have too many ranked outcomes in this dependent variable.
How many outcomes do you have? In ologit, the limit is 50; I don't know about slogit. If you provide some output we might be able to make a better guess.

-------------------------------------------
Richard Williams, Notre Dame Dept of Sociology
OFFICE: (574)631-6668, (574)631-6463
HOME: (574)289-5227
EMAIL: Richard.A.Williams.5@ND.Edu
WWW: http://www.nd.edu/~rwilliam

*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/